Blog · AI and self-service analytics
How to decide what counts as a movement worth reporting for each measure in a weekly sales digest: thresholds from the measure's own week-to-week variability rather than a round number, the floor that stops small denominators generating alerts, the cap on items per digest, and the review that tightens or loosens each threshold from how many alerts were acted on.
A weekly digest that lists eighty movements is not read. One that lists none is not trusted. The difference is the threshold per measure, and most digests set one number for everything. This guide sets out thresholds from each measure's own variability, the denominator floor, the cap, and the quarterly review.
Per measure:
Weekly change = this week − last week Variability = standard deviation of weekly change, trailing 26 weeks Threshold = k × variability, k stated, usually 2 Report if |weekly change| > threshold and denominator ≥ floor
Per digest:
At most n items, largest movements first, n stated, usually 10
| Measure | Level | Variability | Threshold at k=2 | Floor | Alerts last quarter | Acted on |
|---|---|---|---|---|---|---|
| Coverage | Rep | 2.1 pts | 4.2 pts | 10 accounts | 31 | 74% |
| Pipeline coverage | Rep | 0.4× | 0.8× | 5 opps | 58 | 22% |
| Dormant run rate | Region | $18,000 | $36,000 | 9 | 89% | |
| Share of wallet | Segment | 0.9 pts | 1.8 pts | 20 customers | 14 | 36% |
Pipeline coverage generated the most alerts and a fifth were acted on. Its k goes to 3 next quarter. Dormant run rate generated few and almost all were acted on; its k could come down to 1.5 to see whether the extra items are as useful.
Ten items. When more than ten pass their thresholds, the ten largest in value terms are shown and the rest are counted: "and 14 smaller movements". The reader can open them. The digest stays readable.
Quarterly, per measure:
| Acted on | Change |
|---|---|
| Over 70% | Loosen k slightly; the measure may be under-reporting |
| 30% to 70% | Hold |
| Under 30% | Tighten k; the measure is noisy at this threshold |
And a check the other way: movements that were not reported and turned out to matter, found from the account list after the fact. Those tighten the threshold on that measure.
One threshold for all. Silent on the stable measures, constant on the volatile ones.
No floor. The four-account rep is on the digest every week.
No cap. Eighty items; none read.
No review. The thresholds set in month one are the thresholds forever.
Movements without a citation. A threshold breach that does not name the table and row is an alert, not an insight.
Mapped once, the measure history produces the variability, the thresholds and the capped digest every week, and the digest log produces the review every quarter. Covirage builds this from the tables it computes. The sales insights solution describes the setup, and the what makes a change an insight guide covers what the digest says about each movement once the threshold has decided it belongs.
Because five points is a large move in a stable measure and noise in a volatile one. Coverage per rep might move two points in a normal week; pipeline coverage might move twenty. One threshold makes the digest silent on one and constant on the other.
A minimum count beneath the ratio before a movement can be reported: ten accounts, twenty tickets, five opportunities. Below the floor, the measure is shown but not alerted, because one event moves it by a large share.
By a mark on the digest item, opened or dismissed, or by whether the account on the item had a logged activity in the following week. The share of items acted on, per measure, is the digest's own coverage figure, and it drives the review.